We know more about what launched than what lasted.
Three fresh research reviews looked for missing initiatives, challenged the evidence and walked through sharing the report. The next useful step is to follow projects after the announcement.
This review now informs the Presidio profile and disclosure history, the Aramco measurement summary, and the current register exports. It does not add projects to the initiative counts or turn research leads into verified deployments.
The three questions that would make this research much more useful
1. Which pilots became everyday work?
Track the start date, the move into routine use, the eligible wells or sites, and the share actually covered. Record renewals, pauses and retirements. A larger deployment count tells us something; without the eligible population, it cannot tell us how far the rollout has gone.
2. What did the result cost?
Keep production gains, lift-gas savings, staff time and revenue separate. Then ask about integration, hardware, licenses, engineering, support and monitoring. A revenue estimate or a reduction in one operating input is not a net return.
3. What stopped working, and why?
Look for pilots that were stopped, contracts that were not renewed, overrides, false alarms and work people took back. We do not have a reliable failure rate. Public announcements give us a selected sample, and silence after a pilot does not prove abandonment.
Presidio: separate the platform, the result and the next target.
Its August 17 investor deck repeats a 2.3% YTD achieved-uplift claim and sets a 3–5% AI-attributed production-increase goal for 2026. Roughly 2,300 platform wells appears on its goal slide; the active intervention population and coverage date are not established. The goal should sit beside, rather than replace, its earlier quarterly result. Platform coverage is also a different measure from wells receiving an AI-driven operating change.
Our next questions: which wells were eligible, which received an intervention, how was the incremental gain estimated, and what did it cost to keep the system running?
Issuer-reported platform scope and target. The target is not an achieved fleet-wide gain. The current project profile now includes the August 17 disclosure, retaining the earlier quarterly detail separately.
Read the SEC-filed investor presentation · slides 20–23 · Read the existing Presidio profile
A five-well gas-lift trial shows why the mechanism matters.
A June 2026 supplier-authored technical article describes an ORTOmation optimizer on an unnamed Delaware Basin operator’s five-well pad. It adjusted lift-gas setpoints within operating limits over several weeks. The authors report lift-gas use falling from about 4.7 to 2.64 MMscfd, a 44% reduction.
This is a specialist learning and control system. The article does not establish that a language model controlled the wells. It also does not disclose absolute production uplift, full costs or sustained results after the trial. Less lift gas is useful context, but it is not itself a 44% economic return.
What we held back
An anonymous Oklahoma remote-pumper case gives unusual detail about staffing and service pricing. Its page is undated, so we have not admitted it as a confirmed deployment within the review period. Another suggested emissions study still needs a reproducible source check and careful separation of monitoring from AI attribution.
The main register began with a source screen covering 26 September 2024–26 September 2026. Its review date and later supplementary research are separate. Neither the screen nor the subsequent searches establishes complete worldwide coverage.